In this paper, a stochastic model of plug-in hybrid electric vehicles' (PHEV) charging loads is developed in Matlab to investigate its impact on the power grid. In this model, two main types of PHEVs are defined: public transportation vehicles and private vehicles. Different charging time schedules, charging speeds and battery capacities are considered for each type of vehicles. The simulation results reveal that there are two obvious load peaks (at noon and in the evening), if dumb charging method (V0G) is applied. Therefore, smart charging method (V1G) is introduced to mitigate the variation of charging loads. The optimization process is based on the output data of a wind power/battery energy storage system, so as to make better use of surplus wind power. Since variations on both generation and demand sides are decreased, the stability and reliability of the whole power system could be improved dramatically.